发现优质的 AI Agent 技能

**Date:** 2026-02-02 **Tester:** Integration Test Subagent **Agent:** ecap0 **API Base:** https://skillaudit-api.vercel.app

**Date:** 2026-02-02 **Fixed by:** QA Subagent (Claude)

**Datum:** 2025-07-17 **Analysiert:** ferret-scan v1.x (65+ Rules, Regex+AST+Correlation Engine) **Verglichen mit:** ecap-security-auditor (LLM-basierter Audit mit Trust Registry)

**Datum:** 2025-07-17 **Ziel:** Konkrete Patterns aus Ferret-Scan in unseren LLM-basierten audit-prompt übernehmen

> **Date:** 2025-07-14 > **Panel:** Security Auditor · LLM Prompt Engineer · Product Owner > **Problem:** Agent frameworks (llama-index, crewai, autogen) get penalized for patterns that ARE their core purpose

**Reviewer:** Documentation & UX Subagent **Datum:** 2025-07-17

**Date:** 2025-07-13 **Tested Prompt:** `skills/ecap-security-auditor/prompts/audit-prompt.md` **Methodology:** 5 test packages with known ground truth, audited following Steps 1-6 exactly.

**Date:** 2026-02-02 **Tester:** Subagent (adversarial-test) **Skill Version:** Current workspace

This document lists all detection patterns used in AgentAudit security analysis.

This document provides detailed guidance on performing security audits for packages, skills, and MCP servers.

You are reviewing security findings submitted by another agent. Your job is to verify or refute each finding.

You are a security auditor analyzing a software package. Follow the three phases in order. Do not skip phases.